Besides the rapid development of artificial intelligence and computer vision, small and cheap single board computers appeared, which made these technologies available for a much wider audience, so they can be used in e.g. do-it-yourself and hobby projects. Face recognition is one of the most popular topics of computer vision, its appliances ranges from security to entertainment. The presence of these single board computers and computer vision is very prominent if we take a glance at the smart home industry, one can rapidly find many examples for these products and projects.
The basic motivation for my thesis work came from the fact, that there are no open-source, free-to-use modules for single board computers that support the development of applications utilizing face recognition.
During my work I completed three tasks. The first one was to compare the available single board computers’ performance while they are running face recognition algorithms and to determine the key parameters affecting the results. Then I designed a module that helps one to develop applications with such technologies easier and faster. My final task was to develop a proof of concept smart camera application over the module that I have designed.
During the benchmarks I have selected the best performing algorithms and I have elaborated a method based on face tracking that works with sufficient accuracy and is light-weight enough to be used on devices with low computing power. The module and the application are fully functional, while they may have imperfections, and there may be countless ideas for more features. In the future, I would like to continue my work on it, thus make a truly useful module for programming.